aarish22/Algerian_Forest_Fire

# Algerian Forest Fire Prediction This repository contains the code and resources for predicting forest fire indices using various regression models. The dataset used is the **Algerian Forest Fires Dataset**. ## Project Overview The goal of this project is to develop a predictive model for forest fire indices, specifically focusing on the Fire Weather Index (FWI). The dataset includes meteorological data, and the target variable is FWI. ## Dataset The dataset is sourced from Algerian Forest Fires Dataset. ### Data Preprocessing - **Region Classification**: Data was categorized into two regions: `Region 0` and `Region 1`. - **Handling Missing Values**: Rows with missing data were dropped. - **Feature Transformation**: Relevant features were converted to appropriate data types. - **Outlier Detection**: Box plots were used to visualize and handle outliers. ### Target Variable - The target variable is `FWI` (Fire Weather Index). - The `Classes` variable was converted into binary values, representing `Fire` (1) and `Not Fire` (0). ## Exploratory Data Analysis (EDA) - **Correlation Analysis**: A heatmap was generated to identify highly correlated features. - **Distribution Analysis**: Histograms and pie charts were created to visualize data distributions and class proportions. ## Model Training Several regression models were trained on the dataset: 1. **Linear Regression** 2. **Lasso Regression** 3. **Ridge Regression** 4. **ElasticNet Regression** ### Model Evaluation Each model was evaluated based on: - **Mean Absolute Error (MAE)** - **R-squared (R²) Score** Scatter plots of predictions vs. actual values were used to visualize model performance. ## Feature Engineering - **Feature Scaling**: StandardScaler was used to normalize the feature set. - **Correlation Thresholding**: Features with a correlation above 0.85 were dropped to reduce multicollinearity. ## Results | Model | MAE | R² Score | |------------------|-------|----------| | L …